Hi, Sven and PMC members.
First of all, I 100% agree with Miles:
XText has a very good landing page. Excellend, and an example to
follow. As well, the framework
has an excellent reputation, and all of our clients use it with
success.
Sven wrote:
I have to say that I like the text Ed proposed. But I can
imagine (and your analysis certainly makes that clear) that
other members of the EMP have a different view on it. I think
the underlying problem is that there are very different views on
what 'modeling' actually is.
I agree that today, there are different views within the EMP
project, but this is only because we went away from a common vision
which was the strength that brought us to where we are now.
Sven wrote:
I mainly see three different categories:
Sorry for not at all agreeing on this split, I will explain below
why.
Sven wrote:
*OMG's Modeling*
I see all the standards here backed by the ideas around
MDA.
I do no agree to split it. Large parts of the OMG community, and
major industry players (UBS, e.t.c.) agreed that the EMF is a 100%
isomorphic variant of EMof, and thus is the basis for everytthing.
The main idea of the MDA initiative (I cc the main author, so he can
correct me) is to use Mof as the center point, and model all other
standards, methodologies, especially the domain specific ones with
MOF.
EMF is exactly this, with Mof exchanged with ECore. As the OMG
community accepts ECore as the pragmatic implementation of Mof,
Eclipse Modeling Project, with its "onion layers" like Ed Merks
described in his text is thus an a pragmatic implementation of MDA.
Nothing done with ECore can be "non MDA".
Spliting of "OMG's Modeling" from EMP is a huge step back, it throws
us de facto to the wrong thesis that it is simpler to have your own
little DSL than following MDA: EMP as a pragmatic implementation of
MDA has proven the contrary.
Sven wrote:
*Modeling at Runtime*
Using EMF as the domain model for software systems.
CDO is important here, but also things like data binding
and so on.
CDO is an excellent example of a technology that builds
transparently on ECore and thus extends the pragmatic implemenation
of MDA, which EMP is with an important runtime component. CDO shows,
that even things like persistence , that where considered
programmers omain, can be overed with a generic framework, if you
leverage modeling technology.
Sven wrote:
*Modeling as API design on steroids*
Anything that helps building the perfect abstraction for
your problem.
95% of people who know about the real problem in business do not
even know what an API is, but they know there business concepts,
their properties, dependencies, logic. We all worked for decades to
make modeling work for these 95%. As a result, the MDA initiative
found, that doing everything with UML2 is the wrong way, and thanks
to EMP we where able to explore many more ways to express models.
MDA as well found, that the main problem with the 4GL's in the 90is
was not that they could not do anything we can do now, but that the
metamodels of these DSLs where all defined in proprietary, diverse
ways, and thus, all we need is to agree how to define the
metamodesls, and the common agreement is: ECore.
Then thanks to projects like XText and GMF-T we are able to use
these DSLs, without having to create huge commercial frameworks.
I am sorry, but I do not agree here as well: A good problem
abstraction is of course one of the main topics, but this is
certainly not for API design but for
- capturing expert domain knowledge of vertical industries
- abstracting from technical details of common, technical problem
domains for making acccess to underlying implementations easier
- creating simplified domain views, that can be understood by a
larger group of stakeholders.
I respect that there are different views and interests and I
don't think we should try to find an agreement on a single
message.
This is exactly the disaster I try to avoid:
We started from a common vision, we where able to pragmatically
adapt the dogmatic view we need to do all with EMof/CMof to accept a
scaling implementation like ECore as the basis,
and now we give this up.
We had a commonly agreed message in the old text, and if we are not
able to improve it along the lines Miles described, we should simply
keep it.
Did we get an absolute majority of project leaders or committers
vote for chaning the message? Maybe I am too Swiss, but for changing
this we
should have a more democratic process. Fundamental change of the
message without agreement of all stakeholders is not fair.
If some people want to define a new direction, then this is a new
project.
It would just try to be everything to everybody without
attracting or informing anybody in the end. I mainly see two
alternatives
1) We identify the different kinds of 'modeling' (ideally not
more than three) and install a front page dispatching to
individual subpages.
2) We live without a website for EMP and let every project
define its message as it seems fit. E.g. the text proposed by Ed
could become the text for the EMF project website.
100% disagree.
Both alternatives destroy the progress we made, and which was the
basis for a decision to follow the EMP by many big enterprises.
The options I see are
1) We ask for people who find their views not well enough expressed
on the EMP page, and make a link to their pages, telling
that these people define the topic differently than it was up to now
in the EMP.
2) We together agree on an improved version of the old text, which
does NOT change the direction.
For instance, while the text certainly does not enough metion the
work done in the DSL area, it as well doe not clear enough tell that
EMP is commonly agreed to be a pragmatic implemenatation of the MDA,
and that most OMG standards today have a metamodel done in ECore.
3) We commonly agree on a changed view on modeling, which we then
openly express on the EMP page.
I heard and read things like "the frontiers between modeling and
programming are disapearing" and "defining your own DSL and
generating results from it is much a better way than following the
MDA initiative to have a common metamodel of your domain models."
While I do not agree with both, I could life with an open statement
agreed by the EMP community on this.
But simply removing the EMP page, or changing its message
fundamentally does not seem acceptable to me. Of course Microsoft
would be happy to see this, because honestly, and this is only my
personal opinion, the Eclipse Modeling Project, bringing together
the dynamics of open source and open standards is THE killer
argument for the Java platform. Otherwise, if you look just for
cooler, better, smarter software, you may very probably land with
the .net stack.
I am CCing David Frankel, the author of the MDA intiative, Richard
Gronback, who was the co-lead of the EMP with Ed, and had to stop
since Borland stop, and Michael Guttman, the main author of the
Corba standard: three people who invested a lot of their time in
bringing the modeling vision as we agreed on forward, and are
representative of a larger group of people who are relying on the
EMP as a basis to finaly go a big step forward, rather than going
constantly up and down the waves of new framework, new programming
langauges, new platforms.
We would disapoint all of them and destroy a large number of
innovative long-term initiatives by changing the direction.
Ed Merks: I respect your current like for fresh approaches
and clever ideas, as well as real cool things,
but the world relies on you keeping the ship going in the same
direction! It has been a long way.
Regards, Philipp
Dear
Ed and Modeling PMC.
I'd like to comment on the proposed text for the Eclipse
Modling Project.
As I am passionate in the subject, my mail became long, and emotional.
Thus the conclusions ahead:
Ed: as a friend, in my opinion, the new text is a mockery of your oeuvre,
and if you wrote it out of modesty, as I know you, I would propose, to reconsider it.
The old text (see below) was great!
Now below the long analysis.
I wish you all a very nice WE,
Philipp
The new text:
Also as previously discussed, the website is a disaster area. We need a
landing page with a clear message. I propose the following content:
*Modeling: Faster, Smarter, Better*
The bewildering complexity of modern software begs for a fresh
approach focusing on high-level design, delegating menial tasks to
tools and frameworks.From a concise description of your problem
domain, a complete solution can be inferred.
*What is Eclipse Modeling?*
Eclipse Modeling is an integrated assortment of extensible tools and
frameworks for solving everyday problems.
At its core lies the Eclipse Modeling Framework, a rich abstraction
for describing, composing, and manipulating structured information.
Around this core, onion-like technology layers provide powerful
facilities to address most everything you need.
Up to now we had:
"The Eclipse Modeling Project focuses on
the evolution and promotion of
model-based development technologies within the Eclipse
community
by providing a unified set of modeling frameworks,
tooling, and
standards implementations."
If you compare the two versions, and consider that
"Eclipse Modeling (Framework)" is
the name of the project, the word "model" is mentioned
exactly once.
While the old text makes it clear that
- we focus on "model-based" technologies,
- it provides "modeling frameworks and tooling",
- and ends prominently with "standards implementation"
The new text does not even mention that it focuses on
models, it rather mentions as focus
point "high-level design", whereas the specificaiton phase
is not mentioned at all.
The new text tells it provides "an integrated assortment
of extensible tools and frameworks
for solving everyday problems." No mentioning of tooling
and framework to support the
activity of modeling, as in the last text.
Last but not least, the new text presents the project as a
"fresh approach", that is
"Faster, Smarter, Better". It does not even mention, that
we are implementing leading
industry standards like EMof (implemented by ECore),
Mof2Tex (implemented by Acceleo),
QVTO, OCL, UML2, BPMN.
...
